> Fetch clean Markdown by appending `.md` to any page URL under https://signalwire.com/docs or requesting it with the HTTP header `Accept: text/markdown`. The root index at https://signalwire.com/docs/llms.txt lists the available documentation indexes.

# IndexBuilder

> Build searchable vector indexes from document directories.

[sw-search]: /docs/server-sdks/reference/python/agents/cli/sw-search

[documentprocessor]: /docs/server-sdks/reference/python/agents/search/document-processor

[resolvemodelalias]: /docs/server-sdks/reference/python/agents/search/helpers

[buildindexfromsources]: /docs/server-sdks/reference/python/agents/search/index-builder/build-index-from-sources

[buildindex]: /docs/server-sdks/reference/python/agents/search/index-builder/build-index

[validateindex]: /docs/server-sdks/reference/python/agents/search/index-builder/validate-index

`IndexBuilder` processes document directories into searchable indexes. It handles
file discovery, text extraction, chunking, embedding generation, and storage into
either SQLite (`.swsearch` files) or pgvector backends. This is the programmatic
equivalent of the [`sw-search`][sw-search] CLI tool.

```python
from signalwire.search import IndexBuilder
```

> **Warning**
>
> Requires search dependencies. Install with `pip install signalwire-sdk[search-full]`
> for full document processing support.

## **Properties**

**`model_name`** `str`

Name of the sentence transformer model used for embeddings.

---

**`chunking_strategy`** `str`

Active chunking strategy passed to the internal `DocumentProcessor`.

---

**`min_chunk_size`** `int` — default: 0

For the `markdown` strategy, the minimum words a section needs before a
heading is allowed to start a new chunk. Shorter sections merge forward
into the next one instead of being emitted alone. `0` splits at every
heading. Recorded in the index metadata.

---

**`backend`** `str`

Storage backend. Either `"sqlite"` or `"pgvector"`.

---

**`doc_processor`** `DocumentProcessor`

The [`DocumentProcessor`][documentprocessor]
instance used for chunking.

---

**`model`** `SentenceTransformer | None` — default: None

The loaded sentence transformer model. `None` until the first call to
`build_index()` or `build_index_from_sources()`.

---

## **Methods**

#### [build\_index\_from\_sources](/docs/server-sdks/reference/python/agents/search/index-builder/build-index-from-sources)

Build a complete search index from multiple source files and directories.

#### [build\_index](/docs/server-sdks/reference/python/agents/search/index-builder/build-index)

Build a search index from a single directory.

#### [validate\_index](/docs/server-sdks/reference/python/agents/search/index-builder/validate-index)

Validate an existing .swsearch index file.

> Build searchable vector indexes from document directories.

## Docs

- [build_index](https://signalwire.com/docs/server-sdks/reference/python/agents/search/index-builder/build-index.md): Build a search index from a single directory.
- [build_index_from_sources](https://signalwire.com/docs/server-sdks/reference/python/agents/search/index-builder/build-index-from-sources.md): Build a complete search index from multiple source files and directories.
- [validate_index](https://signalwire.com/docs/server-sdks/reference/python/agents/search/index-builder/validate-index.md): Validate an existing .swsearch index file.